[Hawkular-dev] ManageIQ / Hawkular - some questions

[Heiko W.Rupp]

Hi,

sorry for not having responded earlier (and this email
was laying written but not sent in my outbox too :-( )


> Some my preliminar doubts (perhaps this was discussed in a different 
> thread, so please, point me to that if I missed it):
>
> * Where will be stored metrics and inventory ?
>
> That info should live in the provider (owner of that information) and 
> copied to ManageIQ repository ? Or on

Yes. MiQ will pull that via the provider integration code.

> the contrary, ManageIQ will be the owner of that info and the provider 
> is a technical cache.

While it is a cache, it is also the authoritative source.

> * How information of the providers is collected ?
>
> In some preliminary meeting I understood that agents or whatever 
> mechanism that collects information of the middleware domain should be 
> sent to the provider. Is this

Yes
> valid ? or the architecture is that in the future all agents should be 
> managed directly into ManageIQ ?

No.

Please look at figure 1 of
http://www.hawkular.org/blog/2016/02/22/hawkular-manage-iq.html

> I guess that this is some kind of hybrid architecture as there will be 
> always external provider (thinking on Amazon or other domain) that is 
> managed by ManageIQ but the provider is the owner of the info.

Yes.


> model, I'm sure that Amazon cloud will have their own alerting 
> definitions, perhaps an hybrid approach can be used here).

Yes. The question is of course how to drive it. At least
for some of the alerts you need a UI to set them up.
And having 2 places inside the MiQ UI that set up 2
different kinds of alerts could be seen as confusing.

What would certainly good is to be
have "canned" alerts e.g. for resource going up/down,
WildFly needing a reload and a few others where
the notification would result into injecting a MiQ
event into MiQ for further processing.

> * Roles
>
> Are we going to provide strong role profiles ? for example a 
> middleware role, that only can see the Middleware information without 
> any access to cloud/infrastructure.

You mean in MiQ? That needs to be decided.
I think hiding some of their stuff for newbie
users could be helpful.
On the other hand, the knowledge MiQ provides
for the OS-level (be it real, virtual, containerized) and
our knowledge of the application level is a pretty
strong combination.

> How API should work in this case ? All API should be proxied/routed by 
> ManageIQ or accesing to the provider

That is my understanding.

> API is valid ? (Again, thinking on external providers like Amazon 
> makes me wonder what is the goal of this).